iSoMAs: Finding isoform expression and somatic mutation associations in human cancers

TL;DR

The paper introduces iSoMAs, a computational tool that identifies how somatic mutations in cancer affect isoform expression, revealing new gene associations across multiple cancer types.

Contribution

iSoMAs is a novel pipeline using PCA to efficiently detect associations between somatic mutations and isoform-level gene expression across diverse cancers.

Findings

iSoMAs identified 908 genes with mutations significantly associated with altered isoform expression across three or more cancer types.

TP53 mutations were experimentally verified to directly affect isoform expression in cell cycle genes.

iSoMAs outperforms traditional methods with high efficiency and strong biological relevance in pan-cancer analyses.

Abstract

Aberrant alternative splicing, prevalent in cancer, impacts various cancer hallmarks involving proliferation, angiogenesis, and invasion. Splicing disruption often results from somatic point mutations rewiring functional pathways to support cancer cell survival. We introduce iSoMAs (iSoform expression and somatic Mutation Association), an efficient computational pipeline leveraging principal component analysis technique, to explore how somatic mutations influence transcriptome-wide gene expression at the isoform level. Applying iSoMAs to 33 cancer types comprising 9,738 tumor samples in The Cancer Genome Atlas, we identified 908 somatically mutated genes significantly associated with altered isoform expression across three or more cancer types. Mutations linked to differential isoform expression occurred through both cis- and trans-acting mechanisms, involving well-known oncogenes/suppressor genes, RNA binding protein and splicing factor genes. With wet-lab experiments, we verified direct association between TP53 mutations and differential isoform expression in cell cycle genes. Additional iSoMAs genes have been validated in the literature with independent cohorts and/or methods. Despite the complexity of cancer, iSoMAs attains computational efficiency via dimension reduction strategy and reveals critical associations between regulatory factors and transcriptional landscapes.

Somatic single nucleotide variants (SNVs) drive human cancer progression by disrupting alternative splicing (AS), a co-transcriptional process that generates transcript variation and proteome diversity. To better understand the regulatory networks governing splicing, we propose a computational pipeline, iSoMAs (iSoform expression and somatic Mutation Association), and systematically investigate associations between SNVs and isoform expression in a pan-cancer analysis. Our approach leverages principal component analysis (PCA) for dimension reduction in sample-matched mutation and expression data from The Cancer Genome Atlas (TCGA). We identify thousands of genes (termed iSoMAs genes) whose mutation is significantly associated with isoform expression of multiple genes located across different chromosomes in 33 cancer types. Prevalent iSoMAs genes include the well-characterized tumor suppressor TP53 and splicing factor SF3B1. iSoMAs outperforms traditional association study methods with high computational efficiency and stringent control of false positives. Our results show strong biological and clinical relevance, reflecting known and novel functional relationships. We further demonstrate the TP53, R273H mutation’s ability to alter isoform expression by reversing its effects in lung cancer cells. Our findings bring insights into the regulatory networks of isoform splicing and transcription in cancer, bridging genetic and epigenetic regulation of human oncogenesis in an innovative and biologically meaningful way.

Figures